Onion Routing
To protect your identity, NOEMA employs a multi-hop Onion Routing Transport Layer.
The Privacy Dilemma in AI
In traditional Web2 architectures, when you send a prompt to an AI model, you connect directly to a centralized server. That server inherently knows your IP address, your geographic location, and your digital footprint. In a decentralized network, sending your raw IP address directly to a random sovereign GPU operator poses an unacceptable security risk. Your identity must be mathematically separated from your intelligence queries.
How Noema's Relay System Works
NOEMA solves this by ensuring your data never travels in a straight line. Instead, network packets hop through a randomized proxy circuit before they ever reach the destination compute nodes.
By layering the encryption, each node in the circuit only possesses a fraction of the necessary routing information:
- The Entry Relay: This node knows your IP address, but it receives a fully encrypted payload that it cannot read. Its only job is to pass the locked box forward.
- The Middle Relay: This node passes the encrypted data along. It acts as a buffer. It knows neither who you are nor what the destination is.
- The GPU Compute Node (Exit): The sovereign GPU receives the isolated workload to process, but it has zero tracing links back to your originating device.
What This Means for the User
By utilizing this architecture, NOEMA protects you from traffic analysis and metadata harvesting. Even if a malicious actor controls the GPU that is processing your AI workload, it is physically impossible for them to track the workload back to your machine.
When you combine Onion Routing with our local Zero-Knowledge Redaction (ZKR) client, you achieve total anonymity: your IP address is invisible, and your contextual data is masked.